Active Agent pauses & continues for some items causing falase positives

[bbrendon]

Okay. Just caught this agent red handed. Its linux agent v1.1.6.

Connection looks like:
Zabbix Server <--> Internet <--> Firewall/NAT <--> Linux Agent

The internet connectivity didn't go down. This is monitored very closely using two programs.

Here are the results from an agent ping item that runs every thirty seconds.

2007.Aug.14 15:05:23	1
2007.Aug.14 15:04:53	1
2007.Aug.14 15:04:23	1
2007.Aug.14 15:03:53	1
2007.Aug.14 14:47:26	1
2007.Aug.14 14:46:56	1

During the large time gap above, almost all items for the server weren't collecting data. There are about 8 agents in total on the same LAN as this one. None of the other systems triggered as being down.

This happens to other agents as well every so often.

How can I stop stuff like this from happening?

[Alexei]

You may try to increase number of ZABBIX trappers. It may happen that all trappers are busy at some point, so new connections are not accepted.

[JoelG]

trap or ping?

Is it trappers or pingers? I have this exact problem, the interval for pings is anywhere from 30 seconds (what we want) to as much as 3 minutes (not what we want).

I have StartPingers=15, but my StartTrappers=5

2007.Aug.15 12:20:06	Up (1)
2007.Aug.15 12:18:23	Up (1)
2007.Aug.15 12:16:39	Up (1)
2007.Aug.15 12:14:52	Up (1)
2007.Aug.15 12:14:11	Up (1)
2007.Aug.15 12:12:27	Up (1)
2007.Aug.15 12:10:43	Up (1)
2007.Aug.15 12:10:01	Up (1)
2007.Aug.15 12:07:15	Up (1)
2007.Aug.15 12:06:33	Up (1)
2007.Aug.15 12:05:51	Up (1)
2007.Aug.15 12:04:04	Up (1)
2007.Aug.15 12:02:23	Up (1)
2007.Aug.15 12:01:41	Up (1)
2007.Aug.15 11:59:57	Up (1)
2007.Aug.15 11:57:11	Up (1)
2007.Aug.15 11:56:29	Up (1)
2007.Aug.15 11:55:47	Up (1)
2007.Aug.15 11:52:53	Up (1)
2007.Aug.15 11:52:12	Up (1)
2007.Aug.15 11:51:30	Up (1)
2007.Aug.15 11:49:46	Up (1)

[bbrendon]

You may try to increase number of ZABBIX trappers. It may happen that all trappers are busy at some point, so new connections are not accepted.

That was my suspicion as well. After posting I added a simple check item with key value of "tcp,10051" to the host running zabbix_server.

I'm not going to increase trappers at this time. I'm going to wait for this to happen again and hope that this new item correlates with the problem.

Fingers crossed!

[bbrendon]

The simple check mentioned above has been in good status 100% of the time and this has happened a few more times since. I have now added an item: net.tcp.port[x.x.x.x,10051]

This was added to an agent behind the NAT/Firewall. Hopefully this comes up with something.